This document is for general exploration of the India Findex 2014 dataset containing survey results from 3000 participants regarding their financial habits.
## Warning: package 'tibble' was built under R version 3.4.3
## Warning: package 'tidyr' was built under R version 3.4.3
## Warning: package 'forcats' was built under R version 3.4.3
## Warning: package 'knitr' was built under R version 3.4.3
## Warning: package 'data.table' was built under R version 3.4.4
Within academic literature, evidence on the impact on household lifetime wealth of microfinace services is mixed. Some studies such as Pitt and Khandker (1998) found a statistically significant positive relationship between a household’s use of credit and the value of the household assets as well as the likelihood of the children attending school in Bangladesh. However, when Knandker followed up with the subjects in the original review and conducted a second set of surveys to examine borrower behaviour over time, the relationship between credit and markets of household lifetime wealth was greatly diminished. Other studies such as Coleman (1999) examining microcredit in northern Thailand also found no significant impact of credit access to household wealth. However, Cotler and Woodruff (2017) found statistically significant impacts of microlending on the sales and profit margins of small businesses in Mexico.
In order to evaluate the impact of access to microfinace services in isolation, studies must be able to control for sampling bias. Studies such as Karlan and Zinman (2006) manage to do just that by collaborating with a South African MFI to work with randomly chosen participants from a cohort of clients who were on the borderline of loan approval but were rejected. A random selection of these applicants were granted loans and were surveyed both six months and twelve months later to observe the impact of the loan. The researchers found that the chosen borrowers were statistically significantly more likely to retain wage employment, less likely to experience hunger in their household, and less likely to be impoverished than their non-borrowing counterparts.
This report will examine the impact of interacting with financial services on low income households in India. The datasource is the 2014 Global Findex microdata which contains individual-level survey results. Low income households (defined as survey participants in the lower 2 within economy household income quintiles) have been chosen as the focus group as these are the particupants which are most likely to benefit from microfinancial services. The dependent variables are chosen based on their correlations with household lifetime wealth.
The below tables provide a summary of the respondents in the sample. This includes a count of males and females, age groups, the income quintiles and access to bank accounts for the respondents.
Although the distribution of males and females in the sample is relatively proportionate, approximately 150 more males were surveyed in the 2014 dataset in India.
An overwhelming majority of participants had only completely primary education or less (forming 62% of the total sample). Only a further 31% of applications had completed secondary education with only 1% having completed tertiary or higher.
The distrubtion of respondents is relatively even amongst the second, third and fourth quintile. However, respondents in the poorest 20% income bracket are significantly under represented while respondents in the highest 20% income bracket are slightly over represented.
The skew of respondents is right-tailed with many more younger participants than older ones.
The majority of respondents (over 55%) have a bank account.
##Although the global findex database has a wide range of people in the survey, the poorest 20% are under-represented in the sample
Most microfinance institutions are targetting those in the lowest two income quintiles who might otherwise go unserved by traditional financial institutions.
Examining financial decisions made by respondents within the lowest 40% income quintiles, we see the following results.
## [[1]]
## Freq Prop
## Female 484 0.4879032
## Male 508 0.5120968
##
## [[2]]
## Freq Prop
## completed primary or less 705 0.7106855
## completed tertiary or more 31 0.0312500
## secondary 256 0.2580645
##
## [[3]]
## Freq Prop
## 0 513 0.5171371
## 1 479 0.4828629
##
## [[4]]
## Freq Prop
## 0-1 0 0.000000000
## 1-4 0 0.000000000
## 5-9 0 0.000000000
## 10-14 0 0.000000000
## 15-19 132 0.133064516
## 20-24 116 0.116935484
## 25-29 118 0.118951613
## 30-34 110 0.110887097
## 35-39 110 0.110887097
## 40-44 101 0.101814516
## 45-49 78 0.078629032
## 50-54 58 0.058467742
## 55-59 51 0.051411290
## 60-64 56 0.056451613
## 65-69 37 0.037298387
## 70-74 16 0.016129032
## 75-79 6 0.006048387
## 80-84 2 0.002016129
## 85+ 1 0.001008065
Although across the total sample surveyed, 62% of respondents had primary education or less, for those earning income within the lowest 40% quintiles, this is amplied suggesting that low income earners occupy a large subset of those with primary education or less.
On the surface it seems like more people have bank accounts than not
The overall split suggests that the majority of respondents have access to banking.
From the data, we see that examining all respondents suggests that the majority have access to banking services. However, at a closer glance at those in lower incomes this result is flipped. The majority of respondents do not have a bank accout, although the split between those who do and do not is much smaller as compared to the original sample.
The number of survey respondents with bank accounts varies considerably between men and women. In 2014, the majority of female respondents did not have access to a banka ccount. However, an overwhelming majority of males did. This gender disparity can be seen accross the developing world and is statistically significant.
## Warning in `[<-.factor`(`*tmp*`, list, value = "1"): invalid factor level,
## NA generated
## Warning in `[<-.factor`(`*tmp*`, list, value = "1"): invalid factor level,
## NA generated
## [1] NaN
## [1] NaN
## [1] 0.4773869
## [1] NaN
## [1] NaN
Now we shall look at the High Income subsets
## Warning: attributes are not identical across measure variables;
## they will be dropped